Clojure Data Science - Edmund Jackson

Подписаться 29 тыс.

Просмотров 20 тыс.

50% 1

Data science / big data exists at the overlap of traditional analytics and large scale computation. As such, neither the traditional tools of analytics (R, Mathematica, Matlab) nor mainstreams languages (Java, C++, C#) supply its requirements well as they cannot simultaneously provide the mathematical abstractions and real-word platform power that are required. Clojure is privileged in that it has the potential to provide just exactly that. This talk will explore why this is the case, the tools that are available and the challenges that need be overcome for Clojure to realise this potential.

Наука

Опубликовано:

30 июл 2024

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 12

@MaheshCR 11 лет назад

Awesome presentation. Perfect mix of high level overview, zeroing in on a smaller class of problem and not being snobbish. Really liked it.

@u3s1r4 11 лет назад

Great presentation! Found it both thoroughly entertaining and very informative.

@seddona 11 лет назад

Very interesting talk, thanks.

@MaheshCR 11 лет назад

Oh yes Sir, Infinity does occupy my belief system very much, would love the pointers please. Blog post, especially using Clojure, is much appreciated..would be grateful!

@rebcabin 11 лет назад

How to handle zeros in relative entropy? In your example, p=[60 40], q=[100 0], and you compute log(p/q), but there's a divide-by-zero iinm?

@MaheshCR 11 лет назад

Ah..frankly, I did not understand your 3rd sentence onwards and am on my way to Wikipedia right now..thanks for pointing me to something I dont know. Also, you have wisdom to bring humility forward, so any perceived snobbishness serves a purpose. Your comment helped me :)

@thegeniusfool 11 лет назад

If you are interested in infinity in general, I can give some pointers, and I should create a blog post about it, preferably using Clojure to manipulate such transfinite ordinals :-)

@edmundjackson8374 11 лет назад

@David. Oops - you got me ! Thanks for the correction.

@thegeniusfool 11 лет назад

A great presentation. Except for the "x = x + 1" comment regarding infinite numbers. That is obviously not the case even for transfinite ordinals: a successor ordinal is always distinct (and greater in the ordinality sense) from the predecessor, i.e., x+1 != x, even for transfinite numbers. If one interprets "=" as a normative equivalence, i.e., that the cardinality of LHS equals the cardinality of RHS, then his comment is true for transfinite numbers within certain bands only.

@RogerKeulen 11 лет назад

Even the comment's are great: clicking "watch later" now....Brain-Food !

@ggellner 10 лет назад

Sucks he didn't actually test how fast Mathematica is for numerics. Using Entropy instead of Dot[y, Log[y]] seems needless (as this is what he is testing matlab and R for ...). Also odd that he is generating different random numbers for R (between 0.1, 1.0, no 0 and 1.0, maybe R can give back 0?)

@thegeniusfool 11 лет назад

Ugh, right after writing my comment I read yours, and am now sorry for exposing a certain snobbishness in my comment :(